It's important to distinguish two (alleged) purposes of homework: 1) improving learning, and 2) teaching work ethic and organizational skills.
I personally think the second one is a weak argument, until kids are into their teens and learning to organize an open-ended, long-term assignment. For an elementary school kid, I think a much more meaningful lesson is imparted if the kid has to do things they don't like that are ACTUALLY USEFUL, like chores around the house. Doing pedagogically useless homework purely to teach them to do things they don't like seems like an insane distortion of the values it's supposed to impart. Not to mention an egregious intrusion of the schools into the family's role in teaching values.
So, what about the pedagogical value of homework? The 2004 report from Queensland that Val linked is actually wrong when it says that research supports the 10-minutes-per-grade-level rule. That report cites the work of Harris Cooper (Duke University) on this point, and Cooper's own work shows that the evidence for any benefit of homework at all in the elementary grades is exceedingly weak. (In fact, Cooper is more sanguine about the literature than I am. He finds that every single study on this question has flawed methods, but he thinks that on aggregate they add up to something. In fact, you can stack up as many seives as you want, but they're still seives, and the stack is not going to hold water. In particular, not a single study has shown a CAUSAL relationship between homework and performance. Studies do attempt to account for mediator variables -- motivation, parental involvement, socio-economic status, quality of school, quality of teacher, etc. -- but as Cooper says, all of them are flawed. In short, better-prepared students are probably given more homework. The homework is the stone in the stone soup.)
In short, the idea that 10 minutes of homework is the magic right amount for first graders, and 20 minutes for second graders, and so on, is simply pulled out of thin air. It has no empirical support.
